# Free AI Agent, Claude Code Replacement & 7 Other Launches

Channel: The Next New Thing
Video: https://www.youtube.com/watch?v=7TT2aM_DdTk
Duration: 19 min
Language: English
Words: 3940
Transcript page: https://viewrankai.com/tools/youtube-transcript/7TT2aM_DdTk

---

[0:00] Today I got the easiest way to install the most powerful agent on your computer. Even your mom could do it and even your smartest friends are going to love how powerful this is. If you don't want to pay for expensive AI models, I've got a free one that is super powerful that will fit on your computer for free. Ooh, and if you've been using Fable 5 and it's burning up tokens, I've got a plan for you for saving them and a simple tool that will make it easy for your agents to have access to your websites without sharing your passwords with them. All that and so much more

[0:26] news you missed this week. Let's get started. Presented by Zapier, the AI automation company. We got AI releases that you probably missed this week. There's a lot going on. Let's get right into it. This is my absolute favorite one. Hermes is on the desktop. If you wanted an agent and the whole terminal experience felt a little rough for you, this is the one that will make it so accessible that it's like any other app. In fact, I feel like it's like Claude CoWork and I feel comfortable putting it on my main device. Though I have it on a secondary device, too. Here is a quick video

[0:54] showing you what you could do with it. Quen every 20 minutes going and reading Reddit, This is, by the way, his cron jobs. It's so easy now. Cron jobs, I think they should just call it scheduled tasks now because they're making it more user-friendly, right? reading X, searching for other people's challenges. And I'm going to show you what it builds for me here. And I'm going to show you how I take it a step further. Now, if you want to emulate

[1:17] this, you 100% should. You don't need You don't need to do this every 20 minutes if you're using like a cloud. The reason why this works so well with Quen locally is it's free. So, I can have it do it as much as And we've got a uh news story about a free model that you can put on pretty

[1:33] inexpensively. If you're using like ChatGPT or Opus, you'd be paying out the wazoo for this because it costs money every time you do something. So, maybe you do this just once a morning to save a little cash. But, here's the output of that. This cron job. I Why is it taking so long? There we go.

[1:50] That's the output right there. here is You know what? It's taking a little while. So, what I'm going to tell you is he now has a list of problems that people have expressed on Reddit on X and how he could position his company or services to solve those problems and if he's building a company that solves people's problems, he wants to hunt for problems. This is one of the reports that he gets. You've used Hermes. What

[2:13] do you think of them? So the Hermes desktop app literally feels like the Claude desktop app. Just like you said, it feels like a more a much more powerful Claude co-work and what I love about Hermes is you can have one Hermes agent go and spin up additional Hermes agents. So I've had my main Hermes orchestrator go and spin up seven additional Hermes profiles each with their own config, their own tools, their own sole.md, all very specialized, capable of doing very specific things in my business. So I think this is a huge step. Nobody This wasn't going to really take off if it was stuck in the terminal and now I

[2:47] think it will. Yeah, look at this. This is what the desktop app looks like. Everything is labeled so cleanly. You see chat sessions on the left. He's pinned four of them on top. Super clear. If you haven't used it, this is the time to get it. It works really well. Let's go on to the next news story which is of course Claude Fable 5 is out. Karpathy says that this is like having software that comes out of the tap. And I love this guy. Um Here's Here's an example of what that looks like when you can have software

[3:14] just come out of the tap essentially. Claude Fable or Mythos is absolutely insane. On the left here, we have the actual lovable lovable mobile app. And here we have a lovable mobile app that I built in two prompts with Claude Fable 5. And and he by the way is showing not only in this video, not only how the the prompts are stupid simple, just two of them, and then he uses his little lovable clone to build a notion app. It

[3:44] is amazing. People are building with it. What do you think of this? So I've been using it myself. I've actually got it sitting in cursor right here running a slash gold project right now that I spun up this morning. But my initial impressions are it is super powerful. Like it just feels way smarter, way better than me though or than Opus. First thing I did is I said look at my emails, calendar, and call recordings from the past quarter and give me what you think my Q3 priority

[4:13] should be. The priorities it gave me were exactly what I was anticipating, but it went so much deeper and pulled out insights from calls that I barely even remember having that I mean it just absolutely blew my mind with what it gave me. It really feels like a step up from Opus. Great point. You know what? We think about it as something that will create apps for you the way that he did right here in this in this example. Riley is fantastic with it, but I use it even in Claude co-work. It's super simple, but there are problems with it and I want to

[4:44] talk about that. Um Here's one person who says look a $200 Claude Max subscription last 90 minutes using Claude Fable 5. Do not use sub-agents when using Claude Fable 5. So this is draining people's Claude accounts. Here's another one. Claude Fable 5 eats tokens three prompts later and it's 58% of the Max 5X plan. And I want to show this. I'm actually going to zoom in on what JJ said. Jay is at 10X. One of my first interviewees was the founder of 10X. They build they build apps using AI and one of the things that he did was he said look I'm going to show you how much money I spent

[5:20] on it. I'm going to going to fast forward through the good stuff because we've all seen the good stuff. Um and tell you here's what he did or what he says you need to do to avoid burning through tokens. He says run it as a hybrid. He says here's how to keep costs in check without giving up intelligence.

[5:39] Do not run everything in Claude Fable 5. Run a relay across models. Think with Claude Fable 5. Build with 4X or GPT 5.5 or even Sonic 4.6 and then review with Claude Fable 5. I made a mistake yesterday where I said, "You know, this is new. I have to try it. I'm going to have Fable pull all these X posts, all these videos that I'm showing in Co-work." And I did it and it was fantastic. And then the stupid mistake that I made was I had it tweak using Fable. And I would say, "You know what? I want this slide moved here and I want that slide moved there and I

[6:13] want this and that just burned through so much that I kept getting the alert at the end saying, "You're almost out of usage for the week." And so that's that's the big lesson from this. I did the same thing. So, when I was running my initial test with Fable, I was using it in Co-work to analyze my X posts, my LinkedIn posts, repurpose, tell me what I'm doing right, tell me what I'm doing wrong. And I just threw Fable at the whole thing. And like you said, within about 30 minutes, I was at about 70% of not my weekly usage limits, but my session usage limits, which still, with

[6:47] a few prompts, is kind of insane to be burning through that quickly. Um all in all, I think JJ's spot-on. It's like, "Look, you got to We've always talked about knowing when to use the right tool for the job. Like that's always been important. Yeah. But now it's even more important than ever because of how expensive these new models are going to be. And once it's off your subscription, like once you're paying for usage at API rates here in about a week and a half, that's going to

[7:09] be even more important. Yeah, let's get to using it as quickly as possible while it's still available. Okay. This one is also another drop, lesser known, the best open image model, and now it can finally spell. I want to actually not show their intro video, but I found one that's even better. Here it goes. look like. Or here's another example where we can generate a group photo of all these characters. And indeed, most of them do look correct, including Stitch and Doctor Strange, even the

[7:37] Black Panther. Well done. turn Mulan into a man. It's also great at text rendering. So, here's one of my classic prompts and as you can see I'm actually going to stop and I'm going to zoom in because you're not going to see the the benefit of it. Here's what it looks like on the left. It looks nice, but it looks almost like the others until you start paying attention to the details and you see that the first line of the Z image one says oops oops oops oops oops says oh no I'm going to I'm going to zoom out. Uh

[8:02] This is quite a long piece of text. Right. realize oh text is really hard to get right. Here's the one on the right from Ernie it says this is quite a long piece of text and it looks nice until you say you see that it says AI image model could could generate accurately. And so that's the beauty of this model. It works. It's free. It's available right now. Um any thoughts before we move to

[8:26] the next one? Yeah, I mean this has been something that we've seen improving really quickly. I mean think about even a year year and a half ago like pre-nano banana AI images just weren't there and you try to get it to spit out text in an image good luck. I mean it was it couldn't even form words much less spell them correctly. So, it's like we're almost there. I think this is probably the um you know the the breakthrough everybody's been waiting for as far as

[8:49] rendering text in AI generated images. Okay, and people keep asking me for all these links. I'll have a link to everything in fact to this whole report if you want in the description. Next of course Zapier MCP. If you are using Army's agent or any other agent or any other AI tool and you want to give it all your tools, but give it the restrictions that you want to protect your data and the power you need to get it do real work. There's nothing like Zapier MCP. It is the most beautiful interaction where you can just see the tools that you have and then say I want

[9:17] to give it permission to do this not permission to do that. Go use it right now zapier.com/mcp. Next. Meta business agent. I am like super unimpressed with Meta stuff. I think though this is really useful. I'm going to show a video from their launch. I found the right clip. [snorts] Oh wait. Here it is. Because every business in this room and watching online should have an AI agent,

[9:44] full stop. Businesses like Alber Arte, a jewelry brand in Mexico. Owner Veronica Guevara was an early adopter of our business agent and has used it to respond to customers at all hours of the day. So, no message goes unanswered. She tells me that it has saved her 80% of her time. It takes care of all the repetitive customer questions, so she

[10:07] can focus on her craft. You see, with the Meta business agent, you're closing sales even when you're asleep. And I'm excited to announce that the agent is now available across the world to the hundreds of millions of businesses on our WhatsApp Business app and on Messenger. That in- includes right here in the UK. Okay, let me pause here for a second and I'll tell you the cool thing about this is it will not just answer messages, it will also study the way that you responded to your customers in the past and use that to inform how it responds to messages in the future. You can watch it so that

[10:42] you're you're um making sure that it's not doing anything goofy, but it's it seems simple, it works, and I like that they're making this available instantly for people. Thoughts? Yeah, so I think this is a smart move on Meta's part. I mean, obviously WhatsApp is the messaging app of choice for everybody in the world that's not based

[10:58] in the US. So, right. Uh but, you know, what they're talking about there, the capability of responding to customers people that are using WhatsApp for business, I mean, they're they're they're essentially taking the concept that I talk about all the time of speed to lead and baking that into their their native messaging channel. And it's smart. I mean, again, eventually everybody's going to have this. It's just a matter of who gets on it quicker. The fact that it analyzes the way that you already talked to customers and kind of sounds like it's almost building like a brand voice profile for you based on previous conversations is a great feature, and

[11:31] I mean, I'm shocked it took Meta this long, but I think this is going to give them I mean, they have such large distribution with WhatsApp and with Messenger, so I think this is a step in the right direction for them, for sure. Next. Google AI Plus plan is now reduced to $4.99. I put a list together what you get for it. And so, you're getting Gemini usage limits that are increased, Omni flash video, your daily brief that's scheduled, your AI inbox, all this stuff. And still I am yawning. And partially it's because all this stuff kind of comes in the subscription that I

[12:05] already have. And partially because I really don't care about the price. I'm not someone who looks at my Claude monthly subscription at 20 bucks. Actually, it's more than 20, but even my my OpenAI at 20 bucks and goes, "Ah, this is too much." What do you think? I agree, right? I mean, everything that Google's offering here inside this this cheaper plan, like I'm not a price-conscious AI consumer. I think this is geared towards that person that is, you know, they're relatively new to AI, they're in Gmail, they're in Drive, they're in Google Calendar, and they want to kind of dip their toe in the walk in the AI water,

[12:38] so to speak. And this is a really low-risk, low-cost way for them to do that. People like you and I, and for the folks listening to this uh conversation, we're already using Claude or OpenAI or Code X or Claude Code to to do everything that Google is including in that plan there. So, it's like, why would we pay any more, even if it's $5 a month, to do the things that we're already doing with Claude? That's kind

[12:59] of the way that I see it. The only advantage to this one is if you need some people on your team to have a lower price I can't even imagine that. Or people who you know in your life, in your family or work, who are too cheap to get started properly, maybe you give it to them. The other advantage is it maybe creates a little bit of pricing pressure on the the the the companies

[13:18] that we like. All right. Right. Gemini 4. I'm actually going to breeze through this pretty quickly because I the thing that's interesting here is you can now put a full model on your laptop that can do things like we talked Hermes agent. We showed Alex earlier talk about how he gets his to do so much because he's not paying for all these all this access. You can now do that. It is it is amazing. I've got a video here for everyone who's watching. I have another one here that um if you're wondering okay, oh here this

[13:46] is examples of what you can do with it. So these are some examples so we can do like a 7:00 a.m. morning brief. We could say okay, every day brief me on everything that happened and then it just gives us the top priorities, what happened overnight, and what it's going to do today. Here's another example so you could actually triage your inboxes, look at your emails, and that sort of thing. And here's another one with I'll have it for people to look at side by side. He had it oh here this is kind

[14:06] of interesting. review. So we can look at everything inside my memory which we've actually got plugged into the agent operating system as well. So if you go inside the memory section here, we have used Obsidian as a second brain. Okay. It's there for you. It's easy. It's really nice that this is now available and it's coming in smaller and

[14:24] smaller to take on those little tasks. The nice thing about Hermes is it's easy to switch model model. Yeah, I mean the fact that these local models are free. I think this is kind of where everything's going especially as Anthropic like these frontier models get more and more expensive. It's like you don't need a Fable 5 or even like an Opus 48 to do probably 90% of the stuff that the everyday person needs to do with AI. So the fact that these local models are getting um smaller in the sense that like in terms of the actual size and pretty much anybody can put it on their normal hardware, their normal

[14:56] computer, that's going to be the future. I think there's a future where people just are running local models on their machine to do most of their daily lifting and then they'll use a you know a Fable 5 or a GPT 55 or even an Opus to do the heavier task. But most people aren't even going to need that those

[15:13] models. And even for me frankly for most of the things that I do I don't need these heavy models. It's easier to install than I than I realized. I've got this link here so that you can go in and see how to install it. I was actually thinking of walking people step-by-step through his installation. I said, "No, it's a breeze. Just click, click, click, follow instructions, move on." All

[15:28] right. Yep. Here's why I included this. Um I think it's stupid and this is I'm actually going to open this up in a full tweet here. This is OpenAI to me right now. Can you dye my hair to the color of Argentina? And now OpenAI is Dude, what is going on with OpenAI? ChatGPT was number one for a long time. This is the nonsense that they put on and they they post about it. This is not where they their attention should be

[15:59] going. Yeah, the the tweet itself is insane. It's basically saying like, "Hey, use ChatGPT to include a picture of yourself and change the color of your hair." Like, to me they are so it's so mainstream. Like, there there's I think they realize that there's still so many people that have not even used AI and the fact that you can take a picture of yourself and tell it to change the color of your hair is

[16:21] mind-blowing to like 99% of people. They're still targeting that that market, that TAM that's so massive. They, you know, Codex obviously has widespread adoption in the developer community and and people that are advanced are using it for that purpose. But, I think that they're not they're not forgetting the fact that they are like the main like when people think of AI, they think ChatGPT. And I think they're just trying to capitalize on that by by having addressing a as large a TAM as possible. But, I think it's stupid. I mean,

[16:50] I think it's insane. I think it's a mistake. They're constantly telling people they are the Fisher-Price of AI models, that this is the basic stuff. Instead of saying, "We are the high-powered" I'm I'm going to move on. Um I also want to highlight a repo this week. This one has been doing really well this week, getting a lot of

[17:05] attention. It's called Agent Cookie. Here's the problem that it solves. You want to give your agent, even if it's like Claude code or Claude actually not Claude. Claude can actually do really good job in your main browser, but you want to give it access to the sites that you that you need, but you don't want to necessarily log in and so you're sharing cookies, it works really simply. Oh, do I not have the Yeah, here. This is Catherine talking about how she uses it. The sim- the simplicity of being able to give access to your agents, not having them forget how to get in, but not giving them a password,

[17:38] that's the answer here. [snorts] And it's available free right now. Sorry, go ahead. And I saw that the uh I think that repo is posted by Matt Van Horn, who's the same guy behind the last 30 days skill and a lot of other really cool um kind of niche GitHub repos and and tools that I've used and that I know a lot of other people get a ton of value out of. So, if it's coming from him, you know it's going to be high quality and you know

[18:02] it's going to do what it says it does. So, I had no idea he was from him. That's what it said. Yeah, the link the link linked. Yep, M. Van Horn. I've been trying to get him on. He's got to come on here at some point and talk to me. He's been doing incredible stuff. Uh all right. We got links [snorts] to that and everything else, all the links below in the description and uh and now that this is over, I actually have a collection of repos that you can use immediately. There's a link for you

[18:30] right here to watch. Go check that out.

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt